289

A publication of

CHEMICAL ENGINEERING TRANSACTIONS

VOL. 80, 2020 The Italian Association
of Chemical Engineering
Online at www.cetjournal.it

Guest Editors: Eliseo Maria Ranzi, Rubens Maciel Filho

Copyright © 2020, AIDIC Servizi S.r.l.
DOI: 10.3303/CET2080049
ISBN 978-88-95608-78-5; ISSN 2283-9216

Embedding Aspen Custom Modeller for Bioethanol

Fermentation into the Aspen Plus Flowsheet Simulator
Nosaibeh Nosrati-Ghods, Muven Naidoo, Susan T. L. Harrison, Adeniyi J.
Isafiade, Siew Leng Tai*
Department of Chemical Engineering, Faculty of Engineering and the Built Environment, University of Cape Town,
Rondebosch, 7701, South Africa
Siew.tai@uct.ac.za

Commercial processes used to produce first generation biofuels are mature. In this study, a kinetic model
using the equation-oriented process modelling tool Aspen Custom Modeller (ACM) was embedded into the
process simulation environment of Aspen Plus and used to simulate industrial bioreactors with cell recycle for
bioethanol production. To embed a native module in a process simulation environment, the legacy codes
written in ACM were treated as a black box and integrated into the component-based framework. The
compliance of ACM to Aspen Plus allows for proper integration of the reactor simulator developed for the
fermentation process for bioethanol production using kinetic models. The aim of this research is to improve the
ability to use custom kinetic models of fermentation processes, developed in equation-oriented modelling
tools, within a sequential modular flowsheet simulator. The results obtained showed the accuracy of the
selected kinetic model in ACM with less than 1% difference to industrial data, while a stoichiometric model
showed approximately 9% difference to the industrial data. The simulation shows that ACM integrated into
Aspen Plus allows for complex biological processes to be predicted accurately in terms of biomass growth,
ethanol production and sugar consumption.

1. Introduction
Simulation provides a powerful tool to support the design of industrial processes and is now accepted as a
critical step (Rhodes, 1996). Simulation tools for chemical process design can be divided into two categories,
namely sequential modular (or block oriented) flowsheet simulators and equation-oriented process modelling
tools (Schopfer et al., 2004). Sequential modular flowsheet simulator (SMFS) environments (e.g. Aspen Plus,
Aspen HYSYS, and PRO/II) are widely used by process engineers owing to their ease of use and robustness
in handling large-scale process simulation problems (e.g. large number of unit operations, process streams
and chemical components) (Gani et al., 2012). However, most SMFSs are limited in type and number of
process unit operations models and are restricted to continuous steady-state processes (Dimian et al., 2014).
The equation-oriented environments (e.g. gPROMS and ACM) can be used for steady-state and dynamic
process simulation and optimization, as well as for parameter estimation and experimental design (Nawaz,
2015) . Equation-oriented process modelling tools offer the opportunity to develop custom process unit
operation models without the need to develop numerical solution methods for the model equations (Gani et al.,
2012). Therefore, it would be beneficial to combine these two simulation environments. The use of a custom
process unit operation model, developed using equation-oriented modelling tools combined in a SMFS,
requires interfacing between the different software programmes. In the early 1990s, the idea of an open
interface for integration data between process simulation software of various origins was put forward by
academic institutions and industry. Several CAPE-OPEN (Computer Aided Process Engineering) projects
were initiated to develop standards and to explore the possibilities for open interfaces for integration of
process unit operations, thermodynamic and physical property packages, and numerical solvers between the
various process simulation tools (COLaN, 2016). ACM can be used to develop custom process unit operation
models which are not available in the Aspen Plus model library (Brinkmann et al., 2003). After development,

Paper Received: 10 November 2019; Revised: 21 January 2020; Accepted: 18 April 2020
Please cite this article as: Nosrati-Ghods N., Naidoo M., Harrison S.T.L., Isafiade A.J., Tai S.L., 2020, Embedding Aspen Custom Modeller for
Bioethanol Fermentation Into the Aspen Plus Flowsheet Simulator, Chemical Engineering Transactions, 80, 289-294
DOI:10.3303/CET2080049
290

the custom model can be used as a user model block embedded inside a sequential modular flowsheet
simulator (Dimian et al., 2014). For consistent flowsheet simulation and optimization, the use of a custom
model in Aspen Plus requires interfacing with equation-oriented modelling tools (Nikolić, 2016). Testing the
status and performance of the software interoperability, as well as examining the custom model performance
in Aspen Plus with experimental and, specifically, industrial results for full-scale process design would be
essential. Hence, this paper describes the development of the customised kinetic model, its interfacing with
Aspen Plus and the validation with experimental and industrial data for bioethanol production. The current
practice is for simulating a unit operation present in a process flowsheet, in which the required kinetic model of
bioethanol production from sugarcane, considering ethanol inhibition, substrate inhibition and substrate
limitation, does not exist in the model library. To allow a deeper understanding of the impacts of inserting
kinetic models for the fermentation process into the sequential modular flowsheet simulator Aspen Plus,
simulations of bioethanol fermentation process were executed at a small scale and industrial plant scale. A
schematic picture of the unit model interface in Aspen Tech (ACM and Aspen Plus) is shown in Figure 1.
Aspen Custom Modeler (ACM)

Sugarcane Pretreatment Hydrolysis Fermentation Distillation Dehydration

Bagasse

Combined heat
and power
generation

Steam,
electric
energy

Figure 1: The custom model block (ACM) of a process unit operation in a sequential modular flowsheet
simulator (Aspen Plus) for bioethanol production process from sugarcane (Dias et al., 2010)

2. Materials and Methods

2.1 Experimental work for validation
2.1.1 Microbial strains, culture storage and growth media
The strain of Z. mobilis ZM4 (ATCC 31821) was used throughout the study. For short-term storage, all
cultures were stored in the solid rich medium (RM) (Goodman et al., 1982) containing per litre: 20 g glucose,
10 g yeast extract, 2 g KH2PO4, 15 g agar, 1000 ml DI water. Agar plates were stored at 4 °C. For long-term
storage, the cultures were suspended in sterile 40% (v/v) glycerol and stored in 1.5 mL volumes at -60 °C. For
this, equal amounts of the bacterial culture and 80% glycerol were mixed. The cell-glycerol mixture was kept
at room temperature for half an hour prior to freezing.
2.1.2 Inoculum preparation and fermentation studies
Z. mobilis ZM4 (ATCC 31821) was subcultured in fresh inoculum media (10 g L-1 yeast extract, 1 g L-1 MgCl2,
-1 -1 -1
1 g L (NH4)2SO4, 1 g L KH2PO4, with 44 g L glucose) twice before being inoculated into the fermentor. For
all cultivations, the yeast extract and inorganic salts (YEIS) solution were autoclaved separately from glucose.
Inocula were incubated for 24 h at 30 °C in 250 mL conical flasks containing 50 mL medium in a shaker
-1
incubator at a speed of 120 rpm. Initial biomass of 2.1 g L was used to inoculate the fermentor. Initial glucose
-1 -1
concentration and initial ethanol concentration were 48.8 g L and 0.001 g L respectively. Experiments were
conducted in a 7 L BioFlo 110 New Brunswick reactor with a working volume of 5 L in CSTR at an agitation
rate of 200 rpm and 30 °C. The pH was controlled at 6.0 using 1 M NaOH.
2.1.3 Analytical procedures
Dry cell mass from 2 ml samples was determined gravimetrically. The optical density was determined at 660
nm spectrophotometrically (Gensys 10S UV-VIS) and kept at around 0.3 OD by dilution in continuous process.
The fermentation culture was analysed for glucose and ethanol concentrations as a function of time, using
HPLC (Aminex column HPX-87H (300 × 7.8 mm) (Bio- Rad, Ion exclusion column), operated at 65 °C with
-1
filtered (0.45 μm) and degassed 5 mM sulphuric acid as mobile phase at 0.3 ml min , refractive index
detector). Standards containing analytical grade components were used to confirm calibration accuracy.
Samples and standards were filtered with a 0.22 μm syringe filter.
 291

2.2 Computational methods

2.2.1 Model framework
The custom kinetic model for the fermentation process for bioethanol production was developed in ACM (Ver.
9.0) and embedded in Aspen Plus (Ver. 9.0). The model included equations for vapour-liquid equilibrium (VLE)
Eq(1), mass balance Eq(2) and energy balance Eq(3) (e.g. molecular weight, thermodynamic phase equilibria,
kinetic equation). VLE equation (modified Raoult’s law):
= (1)

Mass balance:
+ = + (2)

Energy balance:
+ = + (3)

where Ptot is total vapour pressure (unit pressure), Y is vapour mole fraction (dimensionless), Lmole fraction is
liquid mole fraction (dimensionless), γ is activity coefficient (dimensionless), Pvap is vapour pressure of the
pure component (unit pressure), Z is inlet mole fraction (dimensionless) , F is feed molar flow (mole/unit time),
rrate is reaction rate (mass /unit volume/unit time), Rvol is reactor volume (unit volume), V is vapour molar flow
(mole/unit time), Lmolar flow is liquid molar flow (mole/unit time), Hin is feed enthalpy (energy unit/mole/unit time),
Q is heat (unit energy/unit time), HL is liquid enthalpy (energy unit/mole/unit time) and HV is vapour enthalpy
(energy unit/mole/unit time).
2.2.2 Kinetic models
The kinetic model in this paper incorporates substrate limitation, substrate inhibition, and product inhibition
functions, which are based on modified Monod expressions for glucose conversion using Zymomonas mobilis
ZM4 (Lee and Rogers, 1983; Leksawasdi et al., 2001). The terms are explained in Table 1.
-1
Specific biomass growth rate: Cglu≤100 g L
− (4 )
= , 1− ,
+ −
, , , ,
Specific biomass growth rate: Cglu>100 g L-1
− (4 )
= , 1− , ,
+ − +
, , , , ,
Biomass growth rate:
= (5)
Specific glucose uptake rate:
− (6)
= , 1− , ,
+ − +
, , , , ,
Substrate conversion rate:
= (7)
Specific ethanol production rate:

= (8)
,
Ethanol production rate:
= (9)

2.2.3 Kinetic constants

The kinetic parameters are given in Table 1 for Z. mobilis ZM4 (Lee and Rogers, 1983).
292

Table 1: Kinetics of Z. mobilis ZM4 (Lee and Rogers, 1983)

Constant Description Value
-1
µmax,glu Maximum specific growth rate in glucose (h ) 0.5
-1
Kx,glu Glucose limitation constant for biomass production rate (g L ) 0.5
-1
Pmax,x,glu Maximum ethanol concentration in glucose above which cells do not grow (g L ) 86
-1
Kix,glu Glucose inhibition constant for biomass production rate (g L ) 200-220
Pix,glu Minimum ethanol concentration above which cells production is affected negatively when 22
-1
grown in glucose (g L )
-1
qmax,glu Maximum specific glucose utilization rate (h ) 5.0
-1
Ks,glu Glucose limitation constant for substrate uptake rate (g L ) 0.5
-
Pmax,s,glu Maximum ethanol concentration in glucose above which there is no substrate uptake (g L 127
1
)
-1
Kis,glu Glucose inhibition constant for substrate uptake rate (g L ) 500-2000
Pis,glu Minimum ethanol concentration above which glucose consumption is affected negatively 55
-1
(g L )
Ysp,glu Ethanol yield constant from glucose (g-product/g-glucose) 0.48
µglu, qglu and Specific growth rate on glucose, specific glucose utilization rate
-1
νglu and specific rate of product formation change with time (g L )
Cglu, Cx and Glucose concentration, biomass concentration and ethanol
-1
P concentration change with time (g L )

2.2.4 Stoichiometric model

The ethanol production, biomass growth, sugar consumption, water and carbon dioxide production rates were
calculated by Eqs(10-13) that were sourced from data provided in the Aspen simulation of Brazilian
Bioethanol Science and Technology Laboratory (CTBE) to describe a typical industrial unit (Bonomi et al.,
2011). The by-products (glycerol, and acetic acid) were calculated by Eqs(12-13), thereby reflecting by-
product formation in the real-world processes.
Glucose conversion via complete oxidation Eq(10) and for biomass formation Eq(11):
C6H12O6 → 2 C2H6O + 2 CO2, fractional conversion: 90% of glucose (mole) (10)
3.3 C6H12O6 + 2.8 NH4OH + 2.1 CO2 → CH3COOH + 6.9 H2O+ 19.6 CH1.8O0.9N0.145, fractional (11)
conversion: 100% of NH4OH (mole)
Glucose conversion to by-products: glycerol Eq(12) and acetic acid Eq(13): (12)
0.6 C6H12O6 + 0.5 H2O → 0.5 CO2 + C3H8O3, fractional conversion: 52.09% of glucose (mole)
C6H12O6 → 3 CH3COOH, fractional conversion: 1.315% of glucose (mole) (13)
2.2.5 Mass balance
The mass balance over the CSTR, including the biomass growth rate (rx), substrate uptake rate (rglu), and
ethanol formation rate (re) are given in Eqs(14-16).
= .( 0− ) (14)
= .( 0− ) (15)
= .( 0 − ) (16)
where D is the dilution rate, Cx0 is the initial biomass concentration, Cglu0 is initial glucose concentration, P0 is
the initial ethanol concentration, Cx is the biomass concentration, Cglu is the glucose concentration and P
represents the ethanol concentration.
2.2.6 Process of interfacing
The same versions of process simulation software for the development of the work processes were used in
order to prevent issues of backward compatibility and to retain consistency in the property database version.
The procedures and issues are divided into cases that have to do with physical property interfacing and with
unit model interfacing.

3. Results and discussion

3.1 Comparing kinetic and stoichiometric models with literature experimental data and the
experimental data from this study
The results of the kinetic model developed in ACM and from the stoichiometric model in Aspen Plus in
comparison to the experimental result of this study under anaerobic continuous stirred tank reactor using Z.
-1 -1
mobilis (0.07 h ) for bioethanol fermentation from 48.8 g L of glucose are shown in Figure 2. The kinetic
 293

model showed better agreement with the experimental data than the stoichiometric model (Figure 2). The
ethanol, biomass and residual glucose concentrations from this study were 22.97, 2.94 and 0.0 g L-1
-1
respectively compared to 24.34, 7.95, 0.10 g L from the kinetic model developed in ACM and embedded in
-1
Aspen Plus, and 22.45, 9.79, 4.88 g L from stoichiometric model in Aspen Plus. The calculated regression
2
coefficient (R ) between the kinetic model and the experiment of the study is 0.778 whereas between
stoichiometric model and the experiment of the study is 0.415. Hence, the better agreement was found
between the kinetic model and the experimental data.
concentration (g/l)

concentration (g/l)
20

Residual sugars
4
Ethanol and
biomass

10 2

0 0
Experimental data Kinetic model Stoichiometric model

Figure 2: Comparison of the kinetic and stoichiometric model with experimental data from this study. (|||)
represents ethanol concentration, (///) biomass concentration and (⁞⁞⁞) residual sugars concentration

In another example, the predictions of the kinetic model in ACM and stoichiometric model in Aspen Plus were
compared to the experimental results of Lee et al. (1980) using Zymomonas mobilis for bioethanol production
-1
from glucose at feed concentration of 170 g L ( Figure 3). The results show that the kinetic models gave
better agreement than the stoichiometric model with the experimental results for concentrations of ethanol and
biomass produced, and glucose substrate consumed across dilution rates using a feed substrate
-1
concentration of 170 g.L . The dilution rate, an important factor in CSTRs, played no role when the
-1 -1
stoichiometric model was used and resulted in constant ethanol (78.25 g L ), biomass (11.06 g L ) and
-1
substrate concentrations (170 g L ), as shown in Figure 3.
a) b) c)
Ethanol concentration (g/l)
Glucose concentration

Biomass concentration
(g/l)

(g/l)

0 0.1 0.2 0.3 0 0.1 0.2 0.3 0 0.1 0.2 0.3

Dilution rate (1/h) Dilution rate (1/h) Dilution rate (1/h)

-1
Figure 3: Ethanol production from 170 g L glucose using Z. mobilis (Lee et al., 1980). a) (■) represents
experimental glucose concentration, (---) glucose concentration using stoichiometric model, (—) glucose
concentration using kinetic model; b) (●) represents the experimental ethanol concentration, (---) ethanol
concentration using stoichiometric model, (—) ethanol concentration using kinetic model; c) (▲) represents
the experimental biomass concentration, (---) biomass concentration using stoichiometric model, and (—)
biomass concentration using kinetic model

3.2 Comparing kinetic and stoichiometric models with industrial data

Ethanol production, substrate consumption, by-product production, and microbial growth from glucose during
fermentation were predicted using ACM for the kinetic model and Aspen Plus for the stoichiometric model.
These predictions were compared to the industrial data provided by CTBE (Bonomi et al., 2011). The
conversions for the stoichiometric model were calculated based on industrial data from a typical industrial unit
(Bonomi et al., 2011). The mass fraction ratio of acetic acid over ethanol in industry in Brazil is 3.6 wt%.
Kinetic modelling in ACM embedded in Aspen Plus predicted 3.2 wt% while the stoichiometric model in Aspen
Plus predicted 3.1 wt%. The ratio of glycerol to ethanol in the industry was reported to be 6.33 wt%, lower than
the prediction of 6.9 wt% using the kinetic model and 6.6 wt% using the stoichiometric model. The typical
residual sugar to ethanol ratio in the Brazalian industry was 0.25 wt%, comparable to the ACM kinetic model
prediction (0.3 wt%). The stoichiometric model under-estimated this value at 0.13 wt%. Biomass to ethanol
production ratio in the industry was 5.85 wt%. This was predicted as 4.64 wt% when the kinetic model was
used, but over-predicted at 14.5 wt% by the stoichiometric model. Overall, the differences in results between
industrial data and kinetic model were less than 1 wt% in each case, indicating that the kinetic model
294

developed using ACM embedded into the Aspen Plus framework is an acceptable tool as a simulator of
industrial data. In contrast, the industrial data were not well predicted by stoichiometric conversions across all
components. In particular, the biomass over ethanol fraction was out of range. The stoichiometric equations
were inter-dependent; hence the sensitivity of the conversions affected each component more readily. A small
variation in the conversion led to large misinterpretation of the final outputs.

4. Conclusion
Stoichiometric models can be implemented in sequentially modular simulators with less information and effort.
However, stoichiometric models do not represent continuous experimental data sufficiently because of
uncertainty in the estimation of reaction conversions and the absence of kinetic information to reflect the
impact of dilution rates on CSTR performance. In this study, it was shown that kinetic models developed in
ACM and embedded in Aspen Plus represented experimental and industrial data accurately. Since it is not
possible to develop complex kinetic models of microbial systems in Aspen Plus, these models were developed
using an equation-oriented approach. The custom kinetic model, developed in ACM, served as a case study to
test the current status of software interfaces and custom model performance in Aspen Plus. However, the
interfacing between ACM and Aspen Plus was challenging, hence implementations of proper estimations,
additional scripts and assumptions were required to maximise proper integration between the two platforms.
The complexity of estimating biomass formation remains a key issue in modelling in general. Consideration of
including additional parameters such as by-product inhibition and limitation, and maximum biomass
concentrations in the kinetic model could be considered. However, predicting the behaviour of biological
factors such as cell-to-cell interaction, cell sizes and the cell’s intracellular metabolic state are less accessible
in kinetic models as these remain relatively unknown in mechanism for growth regulation.

Acknowledgment
The financial support of the National Research Foundation of South Africa (CPRR:87744); the research
development grant from the University Research Council (URC) and the Centre for Bioprocess Engineering
Research through the SARChI Chair (GUN:64788) are gratefully acknowledged.

References
Bonomi A., Mariano A.P., Jesus C.D.F., Franco H.C.J., Cunha M.P., Dias M., Chagas M.F., Cavalett O., Mantelatto
P.E., Filho R.M., Junqueira T.L., Cardoso T.F., 2011, Technological assessment program (PAT): The virtual
sugarcane biorefinery (VSB) - 2011 Report.
Brinkmann T., Dijkstra M., Ebert K., Ohlrogge K., 2003, Improved simulation of a vapour permeation module, Journal
of Chemical Technology and Biotechnology, 78, 332–337.
COLaN, 2016, Official website of the CAPE-OPEN laboratories network.
Davis L., Rogers P., Pearce J., Peiris P., 2006, Evaluation of Zymomonas-based ethanol production from a
hydrolysed waste starch stream, Biomass and Bioenergy, 30, 809–814.
Dias M.O.S., Cunha M.P., Jesus C.D.F., Mirna I.G., 2010, Simulation of ethanol production from sugarcane in Brazil:
Economic study of an autonomous distillery, Computer Aided Chemical Engineering, 28, 733–738.
Dimian A.C., Bildea C.S., Kiss A.A., 2014, Integrated process and product design, Computer Aided Chemical
Engineering, 2nd ed. Elsevier B.V., Vol. 35, p. 1-33.
Gani R., Cameron L., Lucia A., Sin G., Georhiadis M., 2012, Process systems engineering, 2. Modeling and
simulation, In Ullmann's Encyclopedia of Industrial Chemistry Wiley-VCH, p. 1–54.
Goodman A.E., Rogers P.L., Skotnicki M.L., 1982, Minimal medium for isolation of auxotrophic Zymomonas mutants,
Applied and Environmental Microbiology, 44, 496–498.
Lee K.J., Rogers P.L., 1983, The fermentation kinetics of ethanol production by Zymomonas mobilis, Chemical
Engineering Journal, 27, B31–B38.
Lee K.J., Skotnicki M.L., Tribe D.E., Rogers P.L., 1980, Kinetic studies on a highly productive strain of Zymomonas
mobilis. Biotechnology Letters, 2, 339–344.
Leksawasdi N., Joachimsthal E.L., Rogers P.L., 2001, Mathematical modelling of ethanol production from
glucose/xylose mixtures by recombinant Zymomonas mobilis, Biotechnology Letters, 23, 1087–1093.
Nawaz Z., 2015, Chemical reactors and process modelling, Oil and Gas Research Journal, 2, 1–2.
Nikolić D.D., 2016, DAE Tools: equation-based object-oriented modelling, simulation and optimisation software,
Computer Science Journal, 2, 1–28.
Rhodes C.L., 1996, The process simulation revolution: thermophysical property needs and concerns, Journal of
Chemical and Engineering Data, 41, 947–950.
Schopfer G., Yang A., von Wedel L., Marquardt W., 2004, CHEOPS: A tool-integration platform for chemical process
modelling and simulation, International Journal on Software Tools for Technology Transfer, 6, 186–202.